1382: A Clinical Decision Support Tool to Prevent Critical Deterioration Events Due to Pediatric Sepsis

Introduction: Delayed recognition and treatment of deterioration is a major patient safety issue. Children who have critical deterioration events (CDEs), defined as a transfer to the intensive care unit and requirement of life-sustaining therapies (e.g., respiratory support, vasoactive infusions) within 12 hours, suffer high mortality rates, incur high healthcare costs, and have prolonged hospital stays. Two major causes of CDEs in children are cardiovascular (CV) and respiratory deterioration secondary to new or unrecognized sepsis. To reduce CDEs at our institution, we developed, validated, and deployed a sepsis prediction model-based clinical decision support (CDS) tool and a “watcher” program. Methods: To assess the effectiveness of our intervention, we compared the rates of CDEs due to CV or respiratory causes in the two years after the implementation of the CDS + “watcher” program to the rates with those in the two-year baseline period. CV CDEs are defined as need for vasoactives within 12 hours of transfer. Respiratory CDEs are defined as need for invasive or non-invasive support within 12 hours of transfer. Results: There were 12,408 patient encounters in the baseline period and 30,745 in the post-implementation period. There were 160 (1.3%) respiratory and 36 (0.29%) CV CDEs in the baseline period and 353 (1.1%) respiratory and 57 (0.19%) CV CDEs in the post-implementation period. The odds of a CV CDE were 37.5% lower in the post-implementation period relative to baseline (OR 0.625, p = 0.03). Patients had no difference in the odds of respiratory events (OR 0.873, p = 0.16). The maximum sepsis score demonstrated strong discrimination for CV events (AUC 0.81 baseline; 0.89 post-implementation) and moderate discrimination for respiratory events (AUC 0.7 in both periods). Conclusions: The sepsis prediction-based CDS tool and “watcher” program were associated with a reduction of CV-related CDEs at our hospital but were not associated with a change in respiratory CDEs, likely due to the sepsis prediction model being tuned to predict primarily the development of septic shock. Further work should focus on deriving a multi-etiology prediction model of CDEs to reduce the rates of other causes of deterioration like respiratory CDEs in hospitalized children.